|
Abstract:
|
Surrogate endpoints are often assessed using a meta-analytic approach. A full surrogate fullfils two conditions: (1) the surrogate is correlated with the clinical endpoint “individual-level surrogacy”), and (2) the effect of treatment on the surrogate predicts the effect of treatment on the clinical endpoint (“trial-level surrogacy”). Biomarker-based surrogates typically fullfil individual-level surrogacy but not trial-level surrogacy. Li and Taylor (2010) have shown that the estimation of treatment effects on clinical endpoints can be enhanced by using biomarkers that are individual-level surrogates, regardless of whether these biomarkers are also trial-level surrogates. Pryseley et al. (2010) have addressed the case of a longitudinally measured outcome, and have shown that under realistic assumptions, a few repeated measurements taken early can adequately predict the outcome and/or treatment effect on the outcome at the final measurement. Actual examples of potential biomarker-based surrogate endpoints will be discussed, including pathological complete response in patients with resectable breast cancer, and prostate-specific antigen in patients with localized prostate cancer.
|
